Introduction to PyTorch

Key concepts and skills covered

• Basics of PyTorch, understanding what tensors are, and how to perform basic operations on them.
• Usage of torch functions like empty, zeros, ones, and randn.
• Understanding of automatic differentiation mechanism provided by PyTorch’s Autograd.
• Understanding the concept of a Linear Layer in a neural network.
• Building a simple linear regression model using PyTorch tensors and Autograd.
• Training a simple linear regression model, including defining a loss function, an optimization function, and updating weights through backpropagation.
• Understanding how to pass data in batches to a model during training.
• Understanding basics of activation functions and how to use them in neural network
• Saving and loading PyTorch models, understanding the difference between saving the whole model and just the state dictionary.
• Understanding the concept of Transfer Learning, how to use pretrained models in PyTorch.
• Learning to modify a pre-trained model for a new task (fine-tuning).
• Recognizing how to use different datasets and how to apply transformations on them using torchvision.transforms.
• Understanding how to evaluate a model by making predictions on test data.
• Simplifying complex neural networks using nn.Sequential in PyTorch.
• Visualizing a neural network
• Visualizing PyTorch model, data and training with TensorBoard

Table of Contents

• Tensor Basics
• Some common PyTorch methods
• Linear Layer
• Automatic differentiation with Autograd
• A Simple Neural Network
• Understanding and Using Activation Functions
• Simplifying Networks with nn.Sequential
• Data Loading and Processing
• Using a Pretrained Network (Transfer Learning)
• Visualizing a neural network
• Visualizing with tensorboard